Annotation of OpenXM_contrib/gmp/mpn/powerpc32/mul_1.asm, Revision 1.1
1.1 ! maekawa 1: dnl PowerPC-32 mpn_mul_1 -- Multiply a limb vector with a limb and store
! 2: dnl the result in a second limb vector.
! 3:
! 4: dnl Copyright (C) 1995, 1997, 2000 Free Software Foundation, Inc.
! 5:
! 6: dnl This file is part of the GNU MP Library.
! 7:
! 8: dnl The GNU MP Library is free software; you can redistribute it and/or modify
! 9: dnl it under the terms of the GNU Lesser General Public License as published by
! 10: dnl the Free Software Foundation; either version 2.1 of the License, or (at your
! 11: dnl option) any later version.
! 12:
! 13: dnl The GNU MP Library is distributed in the hope that it will be useful, but
! 14: dnl WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
! 15: dnl or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public
! 16: dnl License for more details.
! 17:
! 18: dnl You should have received a copy of the GNU Lesser General Public License
! 19: dnl along with the GNU MP Library; see the file COPYING.LIB. If not, write to
! 20: dnl the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
! 21: dnl MA 02111-1307, USA.
! 22:
! 23:
! 24: dnl INPUT PARAMETERS
! 25: dnl res_ptr r3
! 26: dnl s1_ptr r4
! 27: dnl size r5
! 28: dnl s2_limb r6
! 29:
! 30: dnl This is optimized for the PPC604 but it runs decently even on PPC601. It
! 31: dnl has not been tested on a PPC603 since I don't have access to any such
! 32: dnl machines.
! 33:
! 34: include(`../config.m4')
! 35:
! 36: ASM_START()
! 37: PROLOGUE(mpn_mul_1)
! 38: mtctr r5
! 39: addi r3,r3,-4 C adjust res_ptr, it's offset before it's used
! 40: li r12,0 C clear upper product reg
! 41: addic r0,r0,0 C clear cy
! 42: C Start software pipeline
! 43: lwz r8,0(r4)
! 44: bdz .Lend3
! 45: stmw r30,-8(r1) C save registers we are supposed to preserve
! 46: lwzu r9,4(r4)
! 47: mullw r11,r8,r6
! 48: mulhwu r0,r8,r6
! 49: bdz .Lend1
! 50: C Software pipelined main loop
! 51: .Loop: lwz r8,4(r4)
! 52: mullw r10,r9,r6
! 53: adde r30,r11,r12
! 54: mulhwu r12,r9,r6
! 55: stw r30,4(r3)
! 56: bdz .Lend2
! 57: lwzu r9,8(r4)
! 58: mullw r11,r8,r6
! 59: adde r31,r10,r0
! 60: mulhwu r0,r8,r6
! 61: stwu r31,8(r3)
! 62: bdnz .Loop
! 63: C Finish software pipeline
! 64: .Lend1: mullw r10,r9,r6
! 65: adde r30,r11,r12
! 66: mulhwu r12,r9,r6
! 67: stw r30,4(r3)
! 68: adde r31,r10,r0
! 69: stwu r31,8(r3)
! 70: addze r3,r12
! 71: lmw r30,-8(r1) C restore registers from stack
! 72: blr
! 73: .Lend2: mullw r11,r8,r6
! 74: adde r31,r10,r0
! 75: mulhwu r0,r8,r6
! 76: stwu r31,8(r3)
! 77: adde r30,r11,r12
! 78: stw r30,4(r3)
! 79: addze r3,r0
! 80: lmw r30,-8(r1) C restore registers from stack
! 81: blr
! 82: .Lend3: mullw r11,r8,r6
! 83: stw r11,4(r3)
! 84: mulhwu r3,r8,r6
! 85: blr
! 86: EPILOGUE(mpn_mul_1)
FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>